In the current transform SelectFeatures the canonical feature vector for graph representations is x. However, we know that PyG works also with the attribute name edge_attr which encodes edge features. Thus, refining how to encoded "base" or "canonical" representations for different cases should be more intuitive.
In the current transform
SelectFeaturesthe canonical feature vector for graph representations isx. However, we know that PyG works also with the attribute nameedge_attrwhich encodes edge features. Thus, refining how to encoded "base" or "canonical" representations for different cases should be more intuitive.